Closed mkostich closed 9 months ago
BTW, tried w/ https://downloads.pacbcloud.com/public/software/installers/smrtlink_11.1.0.166339.zip
and the equivalent parameters listed by the pbmm2 align --help
there, which differs in that it adds -L 0.5
to --preset ISOSEQ
parameter equivalence list. So command: pbmm2 align -j 16 -k 15 -w 5 -u -o 2 -O 32 -e 1 -E 0 -A 1 -B 2 -z 200 -Z 100 -r 200000 -g 2000 -C 5 -G 200000 -L 0.5 --sort genome.fasta reads.hifi.bam mmap.bam
and, unfortunately, find the same issue. I suspect the Ds vs Ns in CIGAR (mentioned in original issue report) may be a hint. Thanks for any help you can provide.
Why are you trying to manually configure pbmm2
instead of using --preset
? The source code is available, so you can go and figure it out that it's due to this line
Thanks for your help Armin. Users of a pipeline incorporating pbmm2 asked for the option to change some parameters to be able to better match very small exons. They are sophisticated users who need to be able to make adjustments for specific cases. This similar situation can be handled using minimap2 by tweaking settings. So I exposed the pbmm2 mapping parameters in the option file for the pipeline, hoping to enable the same performance, but wanted to keep the default settings for those parameters equivalent to --preset ISOSEQ
, because that is the option I used until receiving the user request. Unfortunately my attempt to set the defaults by following the documentation failed testing. I took a quick look at the code but could not find the right line, and I could not afford to spend too much time looking for it. I hope you don't mind me reporting the mismatch between the documentation and the actual behavior (might benefit others as well), or asking for some guidance. In any case, thanks again for the help. I really do appreciate it. Best regards, Mitch
Operating system
Debian GNU/Linux 11 (bullseye)
Package name
Conda environment
Describe the bug Comparing:
to:
I matched equivalent parameters per the help returned from
pbmm2 align --help
:But I find the results are very different. This should produce a gapped alignment (cDNA vs genomic assembly), and with
--preset ISOSEQ
I see many N (skipped) in cigar string, as expected. MAPQ and edit distance look good. However, with the explicit parameter set (no--preset
used, but the-k 15 -w 5 ...
instead), I see similar CIGAR, but with Ds (deletions) in place of N, much lower MAPQ and much larger edit distances. A typical example of output with--preset ISOSEQ
:While with the explicit parameter list (no
--preset
), the same transcript/genome combo yields:Please note the very similar CIGARs, except the D in place of N in the second case.
Are the parameters I'm passing incorrect or missing something?
Error message Completes without complaint; just many fewer decent hits returned.
To Reproduce Steps to reproduce the behavior. Providing a minimal test dataset on which we can reproduce the behavior will generally lead to quicker turnaround time!
Please try with test data of your choice. I've seen it with various HiFi cDNA datasets, from both mouse and human, so don't think there is anything particular to my data.
Expected behavior A clear and concise description of what you expected to happen.
Expect very similar results, given the equivalency of the
--preset ISOSEQ
option listed bypbmm2 align --help
. Am I missing something?Thanks for your time!